What Engineering Teams Have Found Useful
Feedback from teams who have worked with draviak's guidance library, workshops, and custom engagements.
← Back to HomeWhat Teams Say
The Guidance Library filled a gap we kept running into. We were spending too much time triangulating between vendor documentation and third-party comparisons. Having it structured around the deployment decision itself — rather than the hardware vendor's preferred framing — made it much more usable.
We used the Planning Workshop before finalising our serving infrastructure for a mid-size transformer model. The session pushed us to think through the batch size and latency tradeoff more carefully than we had — the written summary became a reference document for the team. The workshop format was efficient; it didn't drag.
The quarterly updates are the feature we didn't know we needed. The articles from our initial subscription are still accurate on the fundamentals, but the update from Q1 this year pointed out some changes in how certain accelerator categories handle quantised models that were directly relevant to a project we had running.
We commissioned a Custom Engagement when we were consolidating two separate serving pipelines. The discovery session surfaced that we'd framed the problem in a way that was making it look harder than it needed to be. The written guide that came out of it was well-scoped and directly actionable. Delivery was on time.
I was sceptical that a guidance library could add much to what's already available online. But the structure is the thing — the way it maps hardware decisions to model requirements gives you a usable framework rather than a pile of specs to interpret. Worth the subscription for that alone.
The Planning Workshop was useful for our team partly because it brought together engineers who had different assumptions about what our hardware constraints actually were. Working through the framework together surfaced those gaps before we got to procurement. The facilitation didn't feel scripted.
Team Journeys in More Detail
Serving Pipeline Decision for a Document Understanding Model
A Kuala Lumpur fintech team had built a document understanding model and needed to decide between GPU-accelerated cloud serving and CPU-based on-premise infrastructure. Competing internal opinions were stalling the decision.
The team attended a Planning Workshop where the facilitator worked through the model's batch processing pattern, latency requirements, and expected query volume. The workshop identified that the model's characteristics pointed toward CPU-based serving at the team's expected scale.
The team avoided a GPU deployment that would have been over-provisioned for their workload. The written summary became the internal reference document for the infrastructure decision. Timeline: single workshop session plus 2 days for the written summary.
"The workshop gave us a shared vocabulary that we hadn't had before. That was as valuable as the hardware guidance itself." — Infrastructure Lead, Kuala Lumpur fintech
Consolidating Serving Infrastructure Across Two Product Lines
A Cyberjaya-based technology company had two product lines with separate serving infrastructure, each running different model architectures. Leadership wanted to understand whether consolidation was feasible and what hardware configuration it would require.
A Custom Engagement began with a discovery session covering both architectures' requirements and the operational constraints. The engagement produced a structured guide addressing consolidation feasibility, hardware options for the combined workload, and the tradeoffs between a unified and a federated serving approach.
The written guide clarified that partial consolidation was viable for one model pair but not the other, with specific hardware configuration options for each scenario. The walkthrough session allowed the engineering team to ask questions before the infrastructure decision went to procurement. Total engagement: 3 weeks.
"The discovery session redefined how we'd framed the problem. The guide that came out of it was more specific to our situation than anything we could have built from general sources." — Platform Architect, Cyberjaya
Building Team Knowledge Before a Hardware Budget Decision
A Penang-based logistics company's engineering team was preparing a hardware budget proposal for an ML serving deployment. The team needed to understand the hardware landscape well enough to write a defensible proposal, but lacked time for open-ended research.
Two engineers subscribed to the Guidance Library and worked through the relevant reference articles over three weeks. The library's structured organisation meant they could identify which articles applied to their specific model type and serving environment without reading the full catalogue.
The team produced a hardware budget proposal that included a structured comparison of three serving options with specific justification for each. The proposal was approved at the first review. Both engineers retained the subscription for the quarterly updates as their deployment matured.
"The articles gave us the vocabulary and the framework. Writing the proposal became a matter of applying that to our specific numbers rather than trying to synthesise everything from scratch." — Senior Engineer, Penang
Reach the Team
55100 Kuala Lumpur
Sat 10am–2pm MYT
Professional Standing
MDEC Ecosystem Listing
Listed as a relevant B2B information resource in the Malaysia Digital Economy ecosystem directory since 2025.
Asia Cloud & Infrastructure
Associate member of the ACIS technical working group, contributing to regional deployment standards discussions.
Content Review Standards
All library content passes a structured accuracy review before publication, with claims sourced to the relevant hardware generation.
Find Your Route
Talk to us about your team's deployment situation and we'll point you toward the most relevant starting point.
Get in Touch